This systematic review and meta-analysis of randomized controlled trials assessed the effects of exercise on behavioral and psychological symptoms of dementia (BPSD, including depression) in people with dementia (PWD). Secondary outcomes for the effects of exercise were mortality and antipsychotic use. Twenty studies were included in this review (n = 18 in the meta-analysis). Most studies used a multicomponent exercise training (n = 13) as intervention; the control group was often a usual care (n = 10) or a socially-active (n = 8) group. Exercise did not reduce global levels of BPSD (n = 4. Weighted mean difference -3.884; 95% CI -8.969-1.201; I2 = 69.4%). Exercise significantly reduced depression levels in PWD (n = 7). Standardized mean difference -0.306; 95% CI -0.571 to -0.041; I2 = 46.8%); similar patterns were obtained in sensitivity analysis performed among studies with: institutionalized people (p = 0.038), multicomponent training (p = 0.056), social control group (p = 0.08), and low risk of attrition bias (p = 0.11). Exploratory analysis showed that the principal BPSD (other than depression) positively affected by exercise was aberrant motor behavior. Exercise had no effect on mortality. Data on antipsychotics were scarce. In conclusion, exercise reduces depression levels in PWD. Future studies should examine whether exercise reduces the use (and doses) of antipsychotics and other drugs often used to manage BPSD.